Hydroxyl and molecular H₂O diffusivity in a haploandesitic melt

Abstract

H₂O diffusion in a haploandesitic melt (a high-silica and Fe-free andesitic melt, NBO/T = 0.173) has been investigated at 1 GPa in a piston-cylinder apparatus. We adopted a double diffusion couple technique, in which one couple was composed of a nominally anhydrous glass with 0.01 wt.% H₂O and a hydrous glass with 5.7 wt.% H₂O, and the other contained the same nominally anhydrous glass and a hydrous glass with 3.3 wt.% H₂O. Both couples were annealed in a single experimental run and hence experienced exactly the same P–T history, which is crucial for constraining the dependence of H₂O diffusivity on water content. H₂O concentration profiles were measured by both Fourier transform infrared (FTIR) microspectroscopy and confocal Raman microspectroscopy. Nearly identical profiles were obtained from Raman and FTIR methods for profile length >1 mm (produced at 1619–1842 K). By contrast, for profile lengths <100 μm (produced at 668–768 K), FTIR profiles show marked convolution effects compared to Raman profiles. A comparison between the short FTIR and Raman profiles indicates that the real spatial resolution (FWHM) of FTIR analyses is about 28 μm for a 7 μm wide aperture on ∼200 μm thick glasses.

While the short profiles are not reliable for quantitative modeling, the long diffusion profiles at superliquidus temperatures can be fit reasonably well by a diffusivity model previously developed for felsic melts, in which molecular H₂O (H₂O_m) is the only diffusive species and its diffusivity ($D_{{\text{H}}_2{\text{O}}_m}$) increases exponentially with the content of total water (H₂O_t). However, there is noticeable misfit of the data at low H₂O_t concentrations, suggesting that OH diffusivity (D_OH) cannot be neglected in this andesitic melt at high temperatures and low water contents. We hence develop a new fitting procedure that simultaneously fits both diffusion profiles from a single experimental run and accounts for the roles of both OH and H₂O_m diffusion. With this procedure, D_OH/$D_{{\text{H}}_2{\text{O}}_m}$ is constrained to be 0.1–0.2 at 1619–1842 K as H₂O_t concentration approaches zero. The obtained OH diffusivity is similar to fluorine diffusivity but is much higher than Eyring diffusivity.

Introduction

As the most abundant and the most important volatile component in silicate melts, the diffusion of water has drawn substantial attention from volcanologists, geochemists and glass scientists (Shelby, 2008, Zhang and Ni, 2010). Since Zhang et al. (1991) laid out a theoretical basis for water diffusion, especially in demonstrating the dominating role of molecular H₂O (H₂O_m) rather than hydroxyl group (OH), water diffusivity has been experimentally determined for a variety of silicate melts (Zhang and Stolper, 1991, Nowak and Behrens, 1997, Zhang and Behrens, 2000, Freda et al., 2003, Liu et al., 2004, Behrens et al., 2004, Behrens et al., 2007, Okumura and Nakashima, 2004, Okumura and Nakashima, 2006, Ni and Zhang, 2008, Behrens and Zhang, 2009, Ni et al., 2009a, Ni et al., 2009b, Wang et al., 2009, Persikov et al., 2010). It has been generally accepted that in felsic melts, water diffusivity increases strongly with increasing concentration of total water (H₂O_t), but the relationship between diffusivity and H₂O_t concentration is less conclusive for silicate melts of intermediate to mafic compositions.

Water diffusivity in andesitic melts was reported to be nearly independent of water content at superliquidus temperatures of 1558–1848 K (Behrens et al., 2004) and in the intermediate temperature range of 773–948 K (Okumura and Nakashima, 2006). Also at intermediate temperatures, Okumura and Nakashima (2006) found that water diffusivity increases with increasing water content in basaltic and dacitic melts, revealing an unusual behavior of andesite melts. Contrary to previous reports (Behrens et al., 2004, Okumura and Nakashima, 2006), Ni et al. (2009a) and Persikov et al. (2010) demonstrated that, both in the intermediate temperature range and at 1573 K, H₂O_t diffusivity is significantly enhanced by increasing H₂O_t concentration in haploandesitic melts. One may attribute this disagreement to the absence of iron in haploandesitic melts, but in both basaltic (Zhang and Stolper, 1991) and haplobasaltic (Persikov et al., 2010) melts, the increase of water diffusivity with increasing water content appears to be consistent. Whether or not water diffusivity depends on water content has a more profound importance than simply a mathematical curiosity – if H₂O_t diffusivity in andesitic melts at superliquidus temperatures were indeed independent of H₂O_t concentration as previously suggested (Behrens et al., 2004), this would imply that water diffusion in andesitic melts is not dominated by the transport of H₂O_m as in the case of felsic melts (e.g., Zhang et al., 1991, Zhang and Behrens, 2000, Ni and Zhang, 2008, Ni et al., 2009b).

All post-1990 water diffusion studies have used Fourier transform infrared spectroscopy (FTIR), either by measuring a quenched diffusion profile (e.g., Zhang and Behrens, 2000) or by monitoring in situ the change in bulk water content of a dehydrating melt (e.g., Okumura and Nakashima, 2004, Okumura and Nakashima, 2006). FTIR measurement is known for its high reproducibility and sensitivity, and with careful calibration it can also achieve high accuracy. However, due to the limited spatial resolution of FTIR spectroscopy, convolution effects may affect diffusion profiles shorter than 200 μm significantly (Ni and Zhang, 2008). Despite the potential of Raman spectroscopy for measuring H₂O dissolved in glasses, the difficulty in calibrating Raman scattering cross sections has restricted its popularity for many years. In the past decade, increasing needs for analyzing water contents of small melt inclusions have advanced Raman quantification procedures significantly (Chabiron et al., 1999, Thomas, 2000, Behrens et al., 2006, Di Muro et al., 2006, Mercier et al., 2009, Mercier et al., 2010). Nonetheless, the high spatial resolution of Raman microspectroscopy has not yet been applied to study water diffusion, for which spatial resolution is often more critical than the accuracy of measured water contents.

In this study, we conduct experiments at 668–1842 K and 1 GPa to investigate water diffusion in a haploandesitic melt (a high-silica and Fe-free andesitic melt) and analyze the diffusion profiles by both FTIR and Raman microspectroscopy. The emphasis is on the mechanism of diffusion (OH contribution in particular) and the comparison between FTIR and Raman analyses.